Proteolytic stability and aggregation in a key metabolic enzyme of bacteria

Significance

To function properly within a cell, proteins must fold and assemble into their native conformation while avoiding degradation and aggregation. To understand the fundamental relationship between protein assembly, proteolytic stability, and aggregation we harness the lac system of bacteria. This enables us to reversibly perturb assembly of a key metabolic enzyme and quantify its proteolytic stability, propensity for aggregation, and impact on cellular phenotypes. Our work establishes a quantitative experimental system in which genetic and synthetic biology tools can be leveraged to probe and manipulate protein assembly and aggregation, which may inform development of approaches to preventing and treating protein-misfolding diseases.

Abstract

Proteins that are kinetically stable are thought to be less prone to both aggregation and proteolysis. We demonstrate that the classical lac system of Escherichia coli can be leveraged as a model system to study this relation. β-galactosidase (LacZ) plays a critical role in lactose metabolism and is an extremely stable protein that can persist in growing cells for multiple generations after expression has stopped. By attaching degradation tags to the LacZ protein, we find that LacZ can be transiently degraded during lac operon expression but once expression has stopped functional LacZ is protected from degradation. We reversibly destabilize its tetrameric assembly using α-complementation, and show that unassembled LacZ monomers and dimers can either be degraded or lead to formation of aggregates within cells, while the tetrameric state protects against proteolysis and aggregation. We show that the presence of aggregates is associated with cell death, and that these proteotoxic stress phenotypes can be alleviated by attaching an ssrA tag to LacZ monomers which leads to their degradation. We unify our findings using a biophysical model that enables the interplay of protein assembly, degradation, and aggregation to be studied quantitatively in vivo. This work may yield approaches to reversing and preventing protein-misfolding disease states, while elucidating the functions of proteolytic stability in constant and fluctuating environments.

Protein conformational flexibility is intimately connected with both protein aggregation and degradation, as protein unfolding is a key step in proteolytic pathways while conformational rearrangements are involved during nucleation and growth of aggregates (1, 2). Kinetically stable states of proteins, e.g. irreversible transitions during protein folding or assembly, are therefore thought to enhance proteolytic stability and suppress aggregation (3). To elucidate the fundamental principles underlying this relation, its in vivo dynamics, and its cellular outcomes, it is necessary to perturb protein conformational states while assessing the impact on proteolysis, aggregation, and cellular phenotypes. Here, we introduce a bacterial model system with these features by leveraging the lac operon of Escherichia coli and reveal that the key metabolic enzyme β-galactosidase (EC 3.2.1.23; LacZ) is a protein whose assembly process confers proteolytic stability. Disrupting its assembly within cells leads to proteotoxic stress phenotypes, including formation of aggregates, bulk growth rate reduction, and cell death. We unify our findings using a simple biophysical model that enables inference of in vivo rate constants for assembly, degradation, and aggregation.

β-galactosidase is necessary for lactose metabolism in E. coli and is an extremely stable protein. It was previously reported that once lac operon expression has stopped, LacZ levels decreased at a rate of 0.6% per hour, which was within measurement error (4). It was later shown that purified LacZ was not degraded when challenged with various eukaryotic and prokaryotic proteolytic enzymes (5). These findings were further supported by showing that LacZ was not degraded when incubated with an E. coli–specific Lon protease in vitro (6). Our lab has previously shown that LacZ persists in growing cells for multiple generations after expression has stopped, which eliminates lag phases in cyclical fluctuations between glucose and lactose (7). Several functions may be attributed to highly stable proteins such as LacZ (3) e.g., stable proteins can be inherited from mother to daughter cells providing transgenerational memory (7–9), yet the mechanisms underlying the stability of LacZ remain unknown.

Structurally, LacZ is a tetrameric protein (10–19) composed of four identical monomers of 1,024 amino acids (20–23). Monomers consist of five structural domains that include a central triose phosphate isomerase (TIM) barrel. Each active site is formed by amino acid residues from two adjacent monomers across the dimer–dimer interface (24); thus, functional LacZ is tetrameric while free monomers and dimers are catalytically inactive (25–29). The kinetics of association of monomers to form dimers were determined to be reversible while tetramerization is irreversible (27, 28, 30, 31). By screening a library of partial deletion mutants of lacZ, it was found that some LacZ mutants do not tetramerize but rather remain in an equilibrium of monomers and dimers (29, 31). These mutants were missing up to 30 amino acids from the N-terminal end; when supplied in trans with peptides consisting of some or all the missing amino acid residues, LacZ tetramerization occurred and activity was restored (30, 32). The truncated inactive LacZ is referred to as the “α-acceptor” and the N-terminal peptide is called the “α-donor” (29), while the rescue of LacZ activity by combining the two is known as “α-complementation” (29) and underlies the classical blue-white screen of molecular genetics (33).

Previous studies focusing on gene expression costs in the lac operon have shown that addition of an ssrA degradation tag to the LacZ protein leads to reduction in LacZ expression levels (34, 35). These observations indicate that, in certain contexts, native proteolytic pathways can degrade LacZ to some extent, which provides a logical starting point for our investigation of the mechanism of LacZ proteolytic stability. ssrA-mediated protein degradation takes place when ribosomes stall on unstable or untranslatable mRNAs. A tmRNA molecule is used to relieve the stall, and codes for a short 11 amino acid peptide—the ssrA tag—which is attached at the C terminus of the stalled polypeptide marking it for degradation by ClpXP and other proteases (36–39). To facilitate the delivery of ssrA-tagged substrates to ClpXP, the dimeric adaptor protein SspB binds to the first seven amino acids of the ssrA tag, which allows substrate degradation at lower concentrations (39). In recent years, synthetic biologists have used ssrA tags to modulate protein levels without disrupting gene expression (40–43). While extensive work has been performed on different ssrA tags in vitro (36, 44–46), quantification of their in vivo effects on native E. coli proteins is generally lacking.

We address the in vivo proteolytic stability of LacZ, from polypeptide expression to oligomerization of monomers and dimers into functional LacZ tetramers. To probe the mechanism of LacZ’s proteolytic stability, we destabilize LacZ using a set of ssrA tags spanning a range of degradation rates, and use the Miller assay to measure LacZ activity levels (47). LacZ activity measured by the Miller assay and LacZ protein levels measured by western blots were previously shown to be linearly related (48); however, the spectrophotometric Miller assay is substantially more sensitive. We show that despite LacZ-ssrA exhibiting lower expression levels consistent with some degradation, functional LacZ-ssrA is not degraded. We confirm this result in a wide range of ssrA tag configurations, using both the native E. coli ClpXP protease and the nonnative Lon protease of Mesoplasma florum. Using α-complementation we reveal that tetrameric LacZ-ssrA cannot be degraded while its monomeric or dimeric forms can. We demonstrate that disruption of LacZ’s monomers’ and dimers’ assembly into functional tetramers can lead to protein aggregation and cell death. Finally, we combine our findings regarding the mechanism of LacZ proteolytic stability with our measurements of LacZ induction and α-complementation kinetics to develop a simple biophysical model of LacZ tetramerization, degradation, and aggregation. These results provide a detailed mechanistic view of how protein kinetic stability and assembly can be manipulated to impact aggregation and cellular phenotypes and establish a powerful quantitative system for studying these questions.

Results

Attachment of ssrA Tags to the C Terminus of LacZ Does Not Lead to Degradation of Functional LacZ Protein.

We engineered E. coli strains that express the LacZ protein fused to one of six ssrA tags, which were previously determined to span a range of degradation rates when attached to GFP (44–46, 49) (SI Appendix, Table S1). The lacZssrA strain carries the wild-type ssrA tag fused to lacZ, and variant ssrA tags are denoted by superscripts. The ssrA^NYNY and ssrA^NYGSNY tags are known to have the fastest in vitro degradation rates (45). The lacZssrA^LDD strain consists of two amino acid substitutions in the last two amino acids of the ssrA tag that disrupt binding and degradation by ClpXP while keeping the SspB binding site intact (46). We used the ssrA^LDD tag in addition to a wild-type (WT) strain as controls that lack degradation. Editing of the lacZ gene was carried out by lambda red recombineering without leaving genomic scars in the chromosome using the method described in ref. 50.

To evaluate the degradation of LacZ, we measured LacZ activity levels using the Miller assay during growth in glucose and lactose transitions (Fig. 1 A and B). LacZ activity levels were quantified in Miller units (MU), which are proportional to the total LacZ activity of the culture (OD₄₂₀) divided by the cell density (OD₆₀₀) (Materials and Methods). Upon transition to lactose minimal media from glucose, we observed basal LacZ activity levels (20 to 75 MU) among all strains (Fig. 1A). As induction proceeded, strains with different ssrA tags exhibited different LacZ induction profiles. The strongest impact of the ssrA tags was observed in the substantially slower induction of the lacZssrA, lacZssrA^NYNY, and lacZssrA^NYGSNY strains (Fig. 1A). Their steady-state activity levels were 24 to 36% lower than in the WT strain (Fig. 1C) and the difference at mid-induction was even more pronounced, with 43 to 50% lower activity (SI Appendix, Fig. S1 B and C and Table S1). In contrast, the lacZssrA^LDD strain induced nearly as fast as the WT, consistent with the expected lack of degradation in this strain. The remaining two tags, AAV and LAD, showed intermediate induction speeds and steady-state LacZ levels (Fig. 1C and SI Appendix, Fig. S1 B and C and Table S1).

Fig. 1.

Effect of attachment of ssrA tags to the C terminus of LacZ during induction and repression of the lac operon. (A) LacZ activity measured by the Miller assay is shown for WT E. coli and lacZ-ssrA strains in lac operon induction conditions. Cells grown to mid-exponential phase in glucose are switched to minimal lactose media at t = 0. The legend corresponds to all panels. Data are represented as mean ± SEM (n = 3). See also SI Appendix, Figs. S1 A–C. (B) Degradation assay. Cells grown to mid-exponential phase in lactose are switched to minimal glucose media at t = 0. The y-axis is shown on a logarithmic scale. Using linear regression of log activity vs. time the slopes were compared and found to be statistically indistinguishable (P = 0.89, F-test) with a pooled slope value of −0.0043 min⁻¹. (C) Steady-state expression levels of WT and lacZssrA strains. Cells were grown overnight in lactose and then diluted in lactose minimal media and grown to mid-exponential phase. LacZ activity shown as % of WT activity level in steady-state growth conditions in lactose. Bars indicate mean ± SEM (n = 3) and individual replicates are shown as circles. One-way ANOVA and Tukey multiple comparisons test to WT was used for statistical analysis. ****P < 0.0001, ***P < 0.001, **P < 0.01, ns = P > 0.05. (D) Total LacZ activity measurements (OD₄₂₀) of the cells (from Fig. 1B). One-way ANOVA was used to test for differences in total LacZ activity over time, and no significant difference was found (P > 0.05). See SI Appendix, Fig. S1D for OD₆₀₀.

To decouple lac operon expression from LacZ degradation in the ssrA-tagged strains, we performed a degradation assay such that we first grew the strains in lactose, and then switched them at t = 0 to grow in glucose which represses lac expression (Fig. 1B). As expected, the LacZ activity levels were the highest at t = 0 and decayed throughout the experiment, with the WT and the lacZssrA^NYGSNY strains exhibiting the highest and lowest activity levels, respectively, at all time points (Fig. 1B). Surprisingly, however, the rate of exponential decay of LacZ activity was indistinguishable among WT and lacZssrA strains (Fig. 1B, caption). This indicated that if any LacZ degradation were taking place in the lacZssrA strains, it did not have a noticeable effect compared to dilution by cell growth. To determine whether any amount of functional LacZ was being degraded during the experiment, we plotted the total LacZ activity of the cultures (OD₄₂₀) (Fig. 1D). We found that it remained constant in all of the ssrA-tagged strains and did not decay at all while the cells were growing exponentially in glucose (Fig. 1D and SI Appendix, Fig. S1D). From these results, we conclude that functional LacZssrA molecules are not degraded. Consistent with this finding, a quantitative western blot using WT, lacZssrA, and lacZssrA^LDD strains in lactose to glucose growth transition indicated no significant LacZ protein accumulation or reduction over the experiment (SI Appendix, Fig. S1E).

To enhance steric accessibility of the ssrA tags to the SspB adapter protein (44, 45), we attached short flexible peptide linkers consisting of up to three (3) gly-gly-ser-gly repeats between the C terminus of LacZ and the ssrA tag. The degradation assay found no significant change in LacZ levels over the course of the experiment, indicating that even with a more accessible ssrA tag, functional LacZ molecules are not degraded (SI Appendix, Fig. S2).

We tested whether the lack of functional LacZ degradation was specific to the native E. coli ssrA degradation system by using the M. florum Lon protease-ssrA system, which was previously reported to fully degrade LacZ in E. coli (51). We attached the M. florum ssrA tag to the lacZ gene in the chromosome of the WT E. coli and expressed the M. florum Lon protease from a plasmid using an arabinose inducible promoter. Separately, we transformed the WT strain with the Lon protease plasmid as a control. Initially, we tried to recapitulate previous results in which the M. florum Lon protease fully degraded LacZ during lac operon induction with IPTG (Isopropyl β-D-1-thiogalactopyranoside) in LB media (51). We grew the strains to mid-exponential growth in LB media supplemented with 1 mM IPTG and then measured the LacZ activity levels of each strain using the Miller assay (SI Appendix, Fig. S3A). At mid-exponential phase during IPTG induction, the lacZssrA^Mf strain with the Lon protease showed a 55% reduction in LacZ activity relative to the control strain with the Lon protease alone (SI Appendix, Fig. S3A). To ensure Lon protease expression, we performed the LacZ induction measurements at mid-exponential phase in glucose minimal media supplemented with IPTG and increasing concentrations of arabinose (SI Appendix, Fig. S3A). The step increase of the arabinose concentration led to greater decrease in the LacZ activity in the lacZssrA^Mf strain. In contrast, for the control strain lacking the ssrA^Mf tag on lacZ we found no statistically significant change in LacZ activity across all arabinose concentrations. Next, we performed the degradation assay by first growing the two strains in lactose minimal media, and then we transferred the batch cultures to glucose minimal media that was supplemented with increasing arabinose concentrations (SI Appendix, Fig. S3B). Both strains exhibited a growth inhibition in an arabinose dose-dependent manner, which further indicated that the Lon protease was expressed (SI Appendix, Fig. S3C) as previously reported (51). We observed no decay in the total LacZ activity which remained constant throughout the experiment (SI Appendix, Fig. S3D). These results indicate that functional LacZ could not be degraded by the M. florum Lon protease.

GFP but Not Functional LacZ Is Degraded in ssrA-Tagged LacZ-GFP Fusion.

To obtain a positive control for degradation, we used a functional, chromosomally encoded LacZ-GFP fusion protein (52), as GFP was previously shown to be degraded by attachment of ssrA tags (44, 45, 49). In the lacZ-gfp fusion strain, we added the ssrA tag to the C terminus of GFP and followed the LacZ activity in lactose to glucose transitions. At t = 0, we measured the maximal steady-state levels of LacZ among all strains (Fig. 2A), and found that the lacZ-gfp strain had lower LacZ activity compared to the WT strain. We specifically compared the LacZ levels between the lacZssrA and the lacZ-gfpssrA strains to the two control parental strains without the ssrA tag. We found that the addition of ssrA to LacZ-GFP reduced LacZ activity levels by ~30%, similar to the relative reduction we observed for ssrA-tagged LacZ (Fig. 2A and SI Appendix, Table S1). The total LacZ activity remained constant throughout the experiment, indicating that addition of the ssrA tag to a LacZ-GFP fusion did not result in degradation of functional LacZ (Fig. 2B).

Fig. 2.

Degradation assay and microfluidic measurements for strain expressing the ssrA-tagged LacZ-GFP fusion. (A and B) Degradation assay for the lacZ-gfpssrA strain. Cells grown to mid-exponential phase in lactose are switched to minimal glucose media at t = 0. LacZ activity (Miller units) and total LacZ activity (OD₄₂₀) are shown. The legend corresponds to panels (A) and (B). Data are represented as mean ± SEM (n = 3). One-way ANOVA was used to test for differences in total LacZ activity over time [panel (B)], and no significant difference was found (P > 0.05). (C) GFP expression measurements of LacZ-GFP fusion using microfluidics for lac operon induction and repression conditions. Strains expressed ssrA-tagged LacZ-GFP with ssrA, ssrA^LAD, or ssrA^LDD on the 3′ end of the gfp gene. Control strain (lacZ-gfp) lacked an ssrA tag. Cells were grown in lactose and then switched to minimal glucose media at t = 12 h. The dashed line indicates the carbon source perfused shown at right on the Y-axis. The left Y-axis shows GFP expression in arbitrary units (AU). (D) Quantification of the exponential decay of GFP expression after the switch to lac operon repression condition (growth in glucose). Solid lines show fit to a single exponential decay model (circles) corresponding to $y=\left(1-b\right)e^{\left(-kx\right)}+b$. Fitted parameter values: $b$ = 0.057, 0.119, 0.098 (AU), $k$ = 0.325 ± 0.001, 0.567 ± 0.002, 0.350 ± 0.001 (h⁻¹) for lacZ-gfp, lacZ-gfpssrA^LAD, and lacZssrA^LDD respectively. Faster GFP decay represents degradation on top of the normal decay due to cell growth. The Y-axis shows GFP expression in arbitrary units (AU).

To determine whether the ssrA-tagged GFP was degraded when fused to LacZ, we attached two additional ssrA tags (ssrA^LAD and ssrA^LDD) to the C terminus of the LacZ-GFP fusion protein and measured GFP expression in growth transition from lactose to glucose using microfluidics (Fig. 2C). GFP expression was highest in the lacZ-gfp strain, and peaked after 5 h of growth in lactose media, while the lacZ-gfpssrA^LDD strain differed during induction but reached similar levels of GFP fluorescence after 12 h in lactose. In contrast, GFP expression of the lacZ-gfpssrA^LAD strain was about 60% lower than these during growth in lactose, consistent with active GFP degradation during expression. Last, we were not able to measure any GFP expression from the lacZ-gfpssrA strain during growth in lactose, indicating that GFPssrA degradation is much faster than LacZ-GFP expression and maturation. After 12 h of growth in lactose, we switched the growth condition to glucose and measured the exponential decay of GFP expression in each strain (Fig. 2D). We found that both the lacZ-gfp and lacZ-gfpssrA^LDD strains had similar decay profiles, which correspond to the dilution of GFP after lac operon expression stopped. The lacZ-gfpssrA^LAD strain exhibited a significantly faster rate of decay than the lacZ-gfp control strain, corresponding to degradation of GFP in addition to its dilution due to cell growth. Collectively, these results indicate that while the GFP portion of LacZ-GFP is degraded when attached with ssrA tags, functional LacZ is not degraded.

Individual Subunits of LacZ Are Degraded but Tetramerized LacZ Is Protected.

Maturation of the LacZ polypeptide into functional LacZ protein involves several steps, including folding of monomers, dimerization, and tetramerization (24). Since tetramerization has been shown to be irreversible (27, 30), we propose that formation of LacZ tetramers may be the step that confers proteolytic stability. To test this hypothesis, we leveraged α-complementation (29, 33, 53) in which 93 bp were deleted from codon 11 to codon 42 of the lacZ gene (54). This in-phase deletion mutant (ΔM15) maintains an equilibrium of monomers and dimers but cannot tetramerize into functional LacZ protein (31, 55). The LacZ^ΔM15 variant can be complemented through expression of an α-peptide, which spans a portion of the LacZ sequence that contains the ΔM15 deletion (26) and restores LacZ function. In our system, we first deleted the M15 segment of the lacZ gene in the chromosome of both our WT and lacZssrA E. coli strains. We then transformed both strains with the pUC19 plasmid, which contains the α-peptide and was previously shown to complement LacZ^ΔM15 in vivo (56–58). Both the lac operon and the α-peptide are regulated by the lac promoter in these strains, which allows us to induce both using IPTG. To test α-complementation in our engineered strains, we grew the ΔM15 strains with and without the pUC19 plasmid in LB media supplemented with IPTG. We then measured LacZ activity using the Miller assay during 2 h of growth (Fig. 3A). Without the pUC19 plasmid, neither ΔM15 strain showed LacZ activity, as expected. In the pUC19 transformed strains, we measured increasing accumulation of LacZ activity over the course of the experiment. When we compared the LacZ activity of both strains, we measured up to an 84% decrease in LacZ activity in the lacZ^ΔM15ssrA strain compared to lacZ^ΔM15, supporting our hypothesis that monomers and dimers can be degraded. Unexpectedly, we noticed that at t = 0, both strains already displayed substantial LacZ activity prior to the addition of IPTG. These data could be explained by reduced repression by LacI. In this system, both the lac operon in the chromosome and the α-peptide in the high copy number pUC19 plasmid are repressed by a single chromosomal copy of the lacI gene.

Fig. 3.

Using α-complementation to probe mechanism of proteolytic stability of LacZ. (A) α-complementation in lacZ^ΔM15 strains, where LacZ activity is shown for ΔM15 lacZ strains that were transformed with pUC19; controls without pUC19 are shown. Cells were grown to OD₆₀₀ = 0.1 in LB and 1 mM IPTG was added at t = 0. Black and pink triangles overlap. Data indicate mean ± SEM (n = 3). (B) α-complementation in lac operon induction conditions for P_tetlacZ^ΔM15 and P_tetlacZ^ΔM15ssrA strains transformed with pUC19 and pZSint-tetR-lacI. Cells were grown to OD₆₀₀ = 0.15 in LB supplemented with IPTG. At t = 0, 100 ng mL⁻¹ aTc were added. Data indicate mean ± SEM (n = 3). See also SI Appendix, Figs. S4 E and F. (C) α-complementation in lac operon repression conditions for P_tetlacZ^ΔM15 and P_tetlacZ^ΔM15ssrA strains transformed with pUC19 and pZSint-tetR-lacI. Cells were grown to OD₆₀₀ = 0.8 in LB media supplemented with 100 ng mL⁻¹ aTc (t = 0). At t = 0, cells were washed and resuspended in LB media supplemented with 1 mM IPTG. Data indicate mean ± SEM (n = 3). (D) Fold-change in the fraction of LacZ degraded between ΔM15 and parental strains for lac operon induction and repression conditions. (E) Western blot using anti-LacZ antibody for P_tetlacZ^ΔM15, P_tetlacZ^ΔM15ssrA, and P_tetlacZ^ΔM15ssrA^LDD. Strains transformed with pUC19 and pZSint-tetR-lacI were grown to OD₆₀₀ = 0.8 in LB media supplemented with 100 ng mL⁻¹ aTc. At t = 0 cells were washed and resuspended in LB media. The Right panel shows quantification of LacZ^ΔM15 band densitometry. Each time point was normalized to the t = 0 of the P_tetlacZ^ΔM15. Data are represented as mean ± SEM (n = 3).

To observe degradation of LacZ^ΔM15 protein, we sought to decouple lac operon expression from α-peptide production. For this purpose, we replaced the entire promoter of the lac operon in the chromosome with a regulatory module consisting of the tetR gene and tetO promoter, obtained from a Tn10 transposon-containing strain (59). We performed this replacement in the WT, lacZssrA, lacZ^ΔM15, and lacZ^ΔM15ssrA strains. We transformed the four strains with two plasmids, pUC19 and pZSint-tetR-lacI. pZSint-tetR-lacI provides additional copies of the TetR and LacI repressors which in turn provide tighter negative regulation of our engineered lac operon and the pUC19 plasmid respectively. We validated that the promoter replacement worked and that anhydrous tetracycline (aTc) alone, and not lactose or IPTG, induces the lac operon in this strain (SI Appendix, Fig. S4A). We then compared lac operon induction and repression of the P_tetlacZ and P_tetlacZssrA strains by adding or removing aTc. We grew the P_tetlacZ and P_tetlacZssrA strains in LB media supplemented with aTc and followed LacZ activity over 3 h of growth (SI Appendix, Fig. S4B). Under these conditions, the cells expressed LacZ rapidly and reached maximal activity levels that were comparable to the native lac promoter, P_lac (SI Appendix, Fig. S4 A and B). As with P_lac, the attachment of the ssrA tag reduced the LacZ levels by up to 30% compared to the nontagged strain (SI Appendix, Fig. S4B). To verify that we could stop lac operon expression after removing aTc, we grew the same two strains with aTc for 3 h, to an OD₆₀₀ ~0.8, and then centrifuged and washed the cells. After resuspending the cells in media that lacked aTc, we collected samples every 20 min for 2 h while the cells were growing and measured LacZ activity (SI Appendix, Fig. S4C). No statistically significant accumulation of total LacZ activity took place past the washing step (SI Appendix, Fig. S4D).

To measure the dynamics of α-complementation during induction starting from a fully uninduced state and a fully expressed α-peptide, we grew the P_tetlacZ^ΔM15 and P_tetlacZ^ΔM15ssrA strains overnight in LB with IPTG. We then diluted cells into fresh media with IPTG and grew cells to an OD₆₀₀ of 0.15, at which point we added aTc (t = 0). We sampled the growing cultures every 20 min for 3 h of growth and measured their LacZ activity (Fig. 3B). While LacZ activity in P_tetlacZ^ΔM15 continuously increased, activity in P_tetlacZ^ΔM15ssrA increased to only 15% of the levels reached by P_tetlacZ^ΔM15. Plotting the total LacZ activity (OD₄₂₀) vs. the cell density (OD₆₀₀) revealed that α-complementation requires the cells to be at high cell density and therefore lower growth rate (SI Appendix, Fig. S4E). We quantified the efficiency of complementation (SI Appendix, Fig. S4F) and found that at the highest efficiency, 18% of possible tetramers are complemented, which is achieved at OD₆₀₀ ~ 1.1.

To infer the presence of LacZ^ΔM15 protein degradation, we grew the P_tetlacZ^ΔM15 and P_tetlacZ^ΔM15ssrA strains in LB media with aTc for 3 h and then washed and resuspended the cells as above in fresh media lacking aTc. At t = 0, we added IPTG and grew both strains for an additional 2 h while sampling each culture every 20 min followed by LacZ activity measurements (Fig. 3C). Within 40 min after the IPTG addition we observed robust LacZ complementation of the P_tetlacZ^ΔM15 strain. In contrast, the LacZ activity of the P_tetlacZ^ΔM15ssrA strain increased substantially less than in the P_tetlacZ^ΔM15 strain throughout the experiment. To determine how much more degradation takes place in the ΔM15 strains relative to the parental strains, we plot the ratio of the fraction of LacZ degraded between the ΔM15 and parental strains (Fig. 3D). This revealed up to a fourfold increase of the fraction degraded, which increases from ~20% in the parental strain to ~80% in the ΔM15 strain.

To directly observe LacZ^ΔM15 degradation, we performed a quantitative western blot on the ΔM15 strains. We included an additional ΔM15 strain with the control ssrA^LDD tag. We first grew the cells for 3 h in LB media supplemented with aTc to express the lac operon and then washed and resuspended the cultures into fresh media lacking aTc (Fig. 3E). At t = 0, LacZ^ΔM15ssrA protein levels were ~40% of LacZ^ΔM15 levels. Over the course of the 120 min experiment, LacZ^ΔM15ssrA levels were reduced by ~80% relative to initial LacZ^ΔM15ssrA levels. Interestingly, we found a lower but significant reduction of 55% in LacZ^ΔM15 and LacZ^ΔM15ssrA^LDD protein levels over the experiment, indicating that LacZ^ΔM15 is degraded even without the presence of a functional ssrA tag. Taken together, these data demonstrate that disruption of LacZ tetramerization enables proteolytic degradation of LacZ^ΔM15, while tetramerized LacZ is protected from proteolysis.

Proteotoxic Stress Phenotypes Caused by LacZ^ΔM15 Expression.

Previously, it was found that monomers and dimers of LacZ aggregate in vitro (28). To determine whether such aggregation takes place in vivo, we grew P_tetlacZ^ΔM15 and P_tetlacZ^ΔM15ssrA strains with aTc and sampled the cultures in late-exponential growth. To monitor aggregate formation, we stained the cells with both membrane and DNA dyes and visualized the cells using epifluorescence microscopy (Fig. 4A). We found that in late stages of exponential growth, the DNA stain was absent from either the polar or mid-cell regions of the P_tetlacZ^ΔM15 cells. In addition, we found that in some cells the area depleted of DNA stain was surrounded by a membrane (Fig. 4 A and B). In contrast, cells of the P_tetlacZ^ΔM15ssrA strain exhibited a uniform distribution of DNA throughout (Fig. 4A) as did the parental strains lacking the M15 deletion (SI Appendix, Fig. S5A). It was previously shown (60) that inclusion bodies and protein aggregates can be positioned in nucleoid-void regions in the cytoplasm that appear as a DNA stain-free area. In total, we found that in ~40% of the P_tetlacZ^ΔM15 cells DNA staining was indicative of the presence of an aggregate, while the same was found in only ~1% of the P_tetlacZ^ΔM15ssrA cells (Fig. 4C). We note that similar spatiotemporal movement and inheritance of protein aggregates is a well-documented phenomenon in E. coli (60–64).

Fig. 4.

Proteotoxic stress phenotypes of tetramerization-deficient LacZ expressing strains. (A) Fluorescence microscopy of P_tetlacZ^ΔM15 and P_tetlacZ^ΔM15ssrA cells transformed with pUC19 and pZSint-tetR-lacI grown to late-exponential phase in LB media supplemented with 100 ng mL⁻¹ aTc. Cells were stained with FM-4-64 (membrane, magenta) and Hoechst 33342 (DNA, cyan). Arrows indicate positions of aggregates. (Scale bar: 1 µm.) Cells in dashed boxes are shown in panel (B). See also SI Appendix, Fig. S5A. (B) Aggregate surrounded by a membrane in a P_tetlacZ^ΔM15 cell and a P_tetlacZ^ΔM15ssrA cell lacking an aggregate. The dashed line indicates the outline of the membrane stain. (Scale bar: 1 µm.) (C) Frequency of cells with aggregates in P_tetlacZ^ΔM15 (N = 1,733) and P_tetlacZ^ΔM15ssrA (N = 2,566). Data are indicated as mean ± SEM (n = 3). Student’s t test yielded ***P < 0.001. (D and E) Growth dynamics of P_tetlacZ^ΔM15 and P_tetlacZ^ΔM15ssrA cells transformed with pUC19 and pZSint-tetR-lacI over 24 h of growth with 100 ng mL⁻¹ aTc. The legend corresponds to panels (D) and (E). Panel (D) shows cell density measurements (OD₆₀₀) taken every 5 min. Error bars indicate SEM over biological replicates (n ≥ 3). Panel (E) shows instantaneous growth rate measurements of the cultures (from Fig. 4D). Growth rate at time $t$ was calculated as the slope of the $\ln {\mathrm{O}\mathrm{D}}_{600}$ measurements between $t-\mathrm{\Delta }t/2$ and $t+\mathrm{\Delta }t/2$ where $\mathrm{\Delta }t$ = 1 h. See also SI Appendix, Figs. S6 and S7. (F) Quantifying cell death and degradation-mediated rescue in LacZ tetramerization-deficient cells. Cells were grown to late-exponential growth in LB media supplemented with 100 ng mL⁻¹ aTc. Cells were stained with propidium iodide (PI) followed by phase contrast and fluorescence microscopy. (Scale bar: 1 µm.) See also SI Appendix, Figs. S5B, S8 A and B. (G) Single cells showing PI stain penetration into P_tetlacZ^ΔM15 cells and phase dark aggregates (arrows). (Scale bar: 1 µm.) (H) Quantification of positive PI (red) cells as a measurement of dead cells in P_tetlacZ^ΔM15 (N = 1,277) and P_tetlacZ^ΔM15ssrA (N = 1,538). Data indicate mean ± SEM (n = 3). Student’s t test yielded ***P < 0.001. See also SI Appendix, Fig. S8C.

To uncover if other proteotoxic stress phenotypes were associated with aggregates in vivo, we grew the strains with and without aTc, measured their cell densities, and calculated their growth rate dynamics. We found that in the presence of aTc, the P_tetlacZ^ΔM15 strain exhibited a substantial growth rate cost relative to the P_tetlacZ^ΔM15ssrA strain (Fig. 4 D and E and SI Appendix, Fig. S6), while without aTc both strains had similar growth dynamics (SI Appendix, Fig. S7 A–D). The parental strains lacking the M15 deletion exhibited no growth rate differences either with aTc or without (SI Appendix, Fig. S7 E–H). Given the decrease in bulk growth rate observed in P_tetlacZ^ΔM15, we asked whether this strain exhibited an increase in cell death that was associated with the presence of aggregates. We sampled cells from both P_tetlacZ^ΔM15 and P_tetlacZ^ΔM15ssrA strains during late-exponential phase in media that was supplemented with aTc and then stained the cells with propidium iodide (PI) (Fig. 4 F and G). Phase contrast microscopy revealed that, similarly to the formation of aggregates which we observed using fluorescent membrane and DNA stains, there were visible phase dark LacZ aggregates in the P_tetlacZ^ΔM15 cells. Judging by the PI fluorescence within the cells, we found that 22% of the cells were dead (Fig. 4H). In contrast, we observed 2.8% PI-positive cells in the P_tetlacZ^ΔM15ssrA strain. The parental strains lacking the M15 deletion showed normal cell morphology and were PI negative in the same growth conditions (SI Appendix, Fig. S5B). Using Fisher’s exact test we found a significant association (P < 1 × 10⁻¹⁵, two-tailed) between cells containing aggregates and PI-positive cells for the P_tetlacZ^ΔM15 strain (SI Appendix, Fig. S8 D and E). The association was significant even in the P_tetlacZ^ΔM15ssrA strain (P < 2 × 10⁻³, two-tailed).

Modeling Tetramerization, Degradation, and Aggregation Dynamics.

To determine whether the mechanism of proteolytic stabilization of LacZ is consistent with the observed lactose induction kinetics of the WT and ssrA-tagged strains, we fit the measurements using a biophysical model. The model’s dynamical variables are the LacZ monomer ($Z_{\text{mono}}$) and tetramer ($Z_{\text{tetra}}$) concentrations, and its differential equations account for changing growth rates, protein production rates, and degradation during induction on lactose (Fig. 5A and SI Appendix, Table S2 and SI Text). Since cell growth is limited by the carbon flux from lactose, we expected a Monod functional dependence of the growth rate, $\mathrm{\Lambda }$, on the functional LacZ concentration, $Z_{\text{tetra}}$. We fit this dependence with one parameter ($K_{M,\mathrm{growth}}$) using data from the WT and the ssrA-tagged strains (Fig. 5B and SI Appendix, Fig. S9A). To determine the form of the protein production function, we considered LacZ production during and after the lag phase. During the lag phase, in the early stages of lac operon induction, protein production utilizes residual carbon and amino acid pools, and proceeds at a low rate, $k_{\text{basal}}$. After the lag phase, protein production is supported by carbon flux from LacZ, and we fit a Monod function to the production rate (SI Appendix, Fig. S9B).

Fig. 5.

Modeling tetramerization, degradation, and aggregation dynamics. (A) Biophysical model of LacZ induction kinetics, shown separately for lactose induction, which is tied to biomass growth, or induction during α-complementation in LB. Rate constants for each transition are indicated by arrows (see SI Appendix, SI Text for differential equations). (B) Monod relation inferred from induction curves for all strains. The dashed line indicates maximal growth rate ${\mathrm{\Lambda }}_{\text{max}}$ = 0.011 min⁻¹ (measured at mid-exponential phase in lactose). Circles indicate means (n = 3), and bars show ± SEM. Data used from Fig. 1A. The best fit (blue) for a Monod function [${\mathrm{\Lambda }}_{\text{max}}{Z}_{\text{tetra}}/(K_{M,\mathrm{growth}}+Z_{\text{tetra}})$] yields $K_{M,\mathrm{growth}}$ = 730 MU. See also SI Appendix, Fig. S9A. (C) Lactose induction kinetics (points) fit by the model for WT and lacZssrA strains. Data used from Fig. 1A. Inset: Predicted dynamics of LacZ monomers during induction. See also SI Appendix, Fig. S10. (D) Model of induction with α-complementation. Left panel shows $Z_{\text{tetra}}(t)$, Right panels show $Z_{\text{agg}}(t)$ and $Z_{\text{mono}}(t)$; curves correspond to model predictions for lacZ^ΔM15 (black) or lacZ^ΔM15ssrA (pink), with corresponding experimental data from Fig. 3B [circles indicate means (n = 3), bars ± SEM]. The dashed curve shows the model prediction if degradation and aggregation are turned off; the dot-dashed curve has only aggregation off. See also SI Appendix, Fig. S11D.

The duration of the lag phase is determined by both the basal LacZ monomer production rate, $k_{\text{basal}}$, and the tetramerization rate, $k_{\text{tetra}}.$ By numerically solving the model and fitting to the WT induction curve (Fig. 5C), we determined a lower bound for $k_{\text{tetra}}$, and a narrow range of values for $k_{\text{basal}}$ (SI Appendix, Fig. S9E). The model predicts that the concentration of LacZ monomers increases rapidly during the first 20 min of induction, and reaches a steady-state value of $\approx$60 MU, i.e. much below that of LacZ tetramers (Fig. 5 C, Inset). Fixing the parameters inferred from induction of the WT strain, we then fit the induction curves of the ssrA-tagged strains to determine the parameters of ClpXP-mediated degradation (Fig. 5C and SI Appendix, Fig. S9 F and G). The in vivo values that we infer (SI Appendix, Table S2) are consistent with previously known measurements from biochemical assays (Discussion). Overall, the model indicates that the molecular mechanism of proteolytic stability via tetramerization is quantitatively consistent with the measured LacZ induction kinetics (Fig. 5C and SI Appendix, Fig. S10).

We used a modified version of the above model to estimate the rates of α-complementation in the ΔM15 strains, and to assess the impact of aggregated LacZ states on α-complementation dynamics (Fig. 5A and SI Appendix, Table S3 and SI Text). Since induction took place in LB instead of lactose, we used measured growth and production rates as input, and then inferred complementation and aggregation rates by model fitting (SI Appendix, SI Text). As aggregation and growth defects were minimal in P_tetlacZ^ΔM15ssrA (Fig. 4), we used its induction curve to infer the rate of complementation, $k_{\text{comp}}$ = 5.2 × 10⁻³ min⁻¹ (Fig. 5D, pink curve). In the P_tetlacZ^ΔM15 strain, in the absence of aggregation or degradation, the model predicted much more α-complemented induction of LacZ than observed (Fig. 5D, dashed curve). This suggests that in the first half of the experiment ($t\le$ 90 min), nonspecific degradation of LacZ^ΔM15 may slow down α-complementation and delay aggregation. In the second half of the experiment ($t<$ 90 min), even as production decreases substantially (SI Appendix, Fig. S11B), we hypothesize that α-complementation of growing LacZ aggregates speeds up induction. Fitting the model yielded the nonspecific degradation rate and the aggregation rate (SI Appendix, Table S3), and predicted substantial growth of aggregates ($Z_{\text{agg}}$) over the experiment (Fig. 5D, solid black curves). In contrast, for the P_tetlacZ^ΔM15ssrA strain, the model predicts either a negligible or a very low (<200 MU) amount of aggregates depending on whether aggregates can or cannot be degraded (SI Appendix, Fig. S11D).

Discussion

We followed the course of β-galactosidase proteolysis and found that tetramerized LacZ is protected from degradation when attached with ssrA tags (Fig. 1 and SI Appendix, Figs. S2 and S3). We revealed that the mechanism behind this phenomenon involves the reaction of monomers and dimers assembling into functional LacZ tetramers. Prior to tetramer assembly, LacZ monomers and dimers can be degraded as we have shown during lac operon induction and using α-complementation. We found that disruption of LacZ tetramerization leads to proteotoxic stress phenotypes including intracellular aggregates, bulk growth rate reduction, and cell death. Previous studies that attached ssrA tags to LacZ did not come to these conclusions because LacZ levels were observed only during lac operon inducing conditions (34, 35, 51). We went beyond lac operon induction to specifically measure the proteolytic stability of functional LacZ after expression has stopped, which enabled these findings.

To reach our conclusion regarding the lack of degradation of functional LacZ, we attempted to increase the likelihood of degradation in several ways. First, we attached flexible peptide linkers which would increase the accessibility of the ssrA tag to the SspB adapter protein. Even with a more accessible ssrA tag, functional LacZ was not degraded (SI Appendix, Fig. S2). The longer versions of the linker showed a decreased level of degradation which we speculate may be related to higher loss of conformational entropy for longer linker lengths upon binding by SspB. Second, we attached the ssrA tag to the GFP portion of a LacZ-GFP fusion (Fig. 2). The attachment of GFP alone decreases LacZ activity levels which may be attributed to the longer length of the combined polypeptide. The polypeptide is expected to take 1.23 times longer to translate (LacZ-GFPmut2 1,262 AA/LacZ 1,024 AA), which means the rate of production is 1/1.23 = 81% of the WT rate; the relative expression of LacZ-GFP to WT is 84%. Addition of the wild-type ssrA tag to the LacZ-GFP fusion results in efficient degradation of the GFP portion but not of the functional LacZ. Our results support previous findings regarding the fast degradation rate of GFPssrA both in vitro and in vivo (44, 49, 65). The fact that on the same polypeptide GFP is fully degraded while LacZ is not rules out the possibility that ClpXP or SspB is saturated. Specifically, ClpXP/SspB saturation cannot explain the lack of degradation of functional LacZ molecules. The intramolecular nature of this partial degradation is consistent with a mechanical or structural barrier, which when formed prevents ClpXP from degrading the LacZ portion of the polypeptide, and may be related to difficulty in unfolding of certain protein structures (66, 67). Supporting this is the comparable ratio of the LacZ activity of LacZ-GFPssrA to LacZ-GFP (64%) to that of LacZssrA to LacZ (68%), indicating a similar fraction of LacZ-GFPssrA polypeptides are fully degraded. Third, in pursuit of a degradation system that would degrade functional LacZ, we tested the Lon-ssrA system of the M. florum bacteria. It was previously shown that LacZ is fully degraded by the M. florum Lon protease when tagged with ssrA^Mf tag during lac operon induction (51). In the lacZssrA^Mf strain expressing the Lon protease, we found a decrease of LacZ activity of similar magnitude to that observed in ref. 51, indicating that the Mf-Lon system is functioning consistently with previous results; however, our strains have a substantially higher LacZ expression level. Nonetheless, when we attempted to measure LacZ degradation in conditions where Lon-Mf is expressed, functional LacZ was not degraded (SI Appendix, Fig. S3). Despite the fact that Lon-Mf is considered to be a powerful unfoldase (51) it is unable to degrade functional LacZ. Collectively, our attempts to degrade functional LacZ demonstrate that there exists a critical time window for degradation to take place between polypeptide expression and tetramerization into functional protein. We refer to this process as “transient degradation,” in which a protein exhibits two distinct proteolytic states.

By preventing the tetramerization of LacZ using the lacZ^ΔM15 deletion strain, we allowed the equilibrium of monomers and dimers to persist without tetramerizing (Fig. 3). This in turn increased the time interval in which the ClpXP machinery and other proteases were able to degrade LacZ monomers and dimers. The quantitative western blot revealed that LacZ^ΔM15 monomers and dimers are degraded in vivo, which does not appear to have been previously known (68–70). Degradation of the LacZ^ΔM15ssrA was the strongest among the ΔM15 variants, as each LacZ^ΔM15ssrA molecule produced was marked for specific degradation by ClpXP. In contrast, LacZ^ΔM15 and LacZ^ΔM15ssrA^LDD, which do not have a functional ssrA tag, were targeted for degradation by a different, nonspecific process that may involve different proteases. We found that in vivo α-complementation requires high cell density which we interpret as allowing LacZ^ΔM15 and the α-peptide to reach sufficiently high intracellular concentrations for the complementation reaction to proceed. We speculate that the reason for the low (18%) efficiency of α-complementation is related to α-peptide degradation (6) and the LacZ^ΔM15 degradation revealed here.

Our findings support a simple biophysical model with three molecular species: monomeric, tetrameric, and aggregated LacZ (Fig. 5A). The parameters that we inferred from fitting this model to the in vivo data were consistent with prior biochemical measurements as follows. We inferred the maximal production rate of LacZ monomers to be 57 MU/min or 9.5 monomers per second. For comparison, Kennel and Riezman’s direct measurements found a production rate of 20 LacZ monomers per second, in conditions with faster growth rate (doubling time = 48 min) and induction using IPTG instead of lactose (71). In our model, the duration of the lag phase in the WT strain (Fig. 5C) is determined by the basal production and tetramerization rates. By fitting the WT measurements, we could infer that upon transition to lactose the basal production rate is 12 times lower than the maximal production rate, and the tetramerization rate $k_{\text{tetra}}$ is at least 1 min⁻¹. The biochemical mechanism of tetramerization has been studied in ref. 28 and involves multistep kinetics with two rate-limiting steps: bimolecular reaction of monomers to form dimers, followed by a unimolecular transition of dimers into primed dimers that rapidly associate to form tetramers. The bimolecular and unimolecular rate constants were measured as k₂ = 4.3 × 10³ M⁻¹ s⁻¹ and k₁ = 0.5 × 10⁻³ s⁻¹ (28), respectively, hence once a concentration of k₁/k₂ = 0.1 μM of LacZ monomers has been reached, the rate-limiting reaction is unimolecular. Using 2 nM = 1 molecule per E. coli cell, this corresponds to 50 LacZ monomers, which would be produced within ~1 min at the basal production rate. This estimate supports our simple model of tetramerization as a unimolecular reaction (SI Appendix, SI Text). We note, however, that the rate k₁ measured in vitro is 20 times slower than our minimal inferred tetramerization rate, which may be related to the biochemical unfolding and refolding conditions that were used to set up the reaction and the lower temperature (20 °C), or additional interactions that exist in vivo [e.g., a study of cotranslational tetramerization found substantially higher rates of LacZ tetramerization (72)]. Indeed, if in vivo tetramerization proceeded at the rate k₁, it would take over 30 min for each LacZ tetramer to mature, which is approximately the duration of the lag phase (7).

The degradation rate that we infer for the lacZssrA strain (0.50 min⁻¹) is comparable to in vitro measurements of ClpXP/SspB-mediated degradation of eGFPssrA (73), which determined the Michaelis–Menten constants for degradation, K_M = 0.47 μM and V_max = 0.085 μM min⁻¹, yielding V_max/K_M = 0.18 min⁻¹. Our inference of degradation rates is a lower bound, as it is tied to the tetramerization rate in the model, which we set at the minimum inferred value. If we assume a higher tetramerization rate, the window for transient degradation becomes shorter, and a higher degradation rate would be inferred. As this would deviate further from the in vitro measurement, a unimolecular tetramerization rate of ~1 min⁻¹ appears to be a reasonable assumption. In LB, we inferred a somewhat lower LacZssrA degradation rate (0.3 min⁻¹) which could indicate a higher burden on the ClpXP machinery to support rapid growth. For the LacZ^ΔM15 protein, we inferred a nonspecific degradation rate of 0.12 min⁻¹, corresponding to a half-life of 5.8 min. This value is comparable with previous measurements on a library of truncated LacZ proteins which exhibited in vivo half-lives ranging from minutes up to an hour (69). It is important to note, however, that these truncated LacZ proteins were not functional and could not be α-complemented.

Our visualization of aggregation in vivo revealed that in a subset of cells, the aggregates were surrounded by a membrane. We speculate that aggregates may sterically exclude the nucleoid leading to a nucleoid-free space, which enables membrane invagination through ectopic engulfment by the divisome (62, 74). Given that aggregation involves a nucleation process (75–77), the inheritance of aggregates increases the probability of aggregate growth in subsequent generations. The nonspecific degradation of LacZ^ΔM15 is evidently not sufficient to prevent aggregation (Fig. 4), however by attaching an ssrA tag to LacZ^ΔM15, aggregate formation was minimized. The fact that i) only LacZ has been perturbed in the lacZ^ΔM15 strain, ii) specific ssrA-mediated degradation of LacZ^ΔM15 reverses the aggregation phenotype, and iii) LacZ monomers/dimers aggregate in vitro strongly supports our inference that LacZ^ΔM15 aggregates in vivo. Nevertheless, it is possible that the composition of the aggregates we observe contains additional proteins other than LacZ^ΔM15. Our biophysical model recapitulated these results, predicting that aggregates grow to ~1,500 MU (~3,750 tetramers per cell) during a typical induction experiment (Fig. 5D), while ClpXP-mediated degradation strongly reduces aggregation (SI Appendix, Fig. S11D). Based on the volume of a LacZ tetramer in the crystal structure (~2.1 × 10³ nm³) (24), the linear dimension of an aggregate would be approximately 200 nm, which places a lower bound on the expected in vivo size, and is consistent with detection by light microscopy (Fig. 4G).

Our bulk growth rate measurements revealed a pronounced growth rate reduction in the P_tetlacZ^ΔM15 strain that was abolished in P_tetlacZ^ΔM15ssrA (Fig. 4E). To test whether there is a direct connection between aggregates and growth phenotypes at the single cell level, we quantified a large number of cells for presence or absence of aggregates, and presence or absence of PI strain, a marker of cell death. This analysis revealed that presence of aggregates is associated with cell death via Fisher’s exact test (SI Appendix, Fig. S8 D and E), and the association was highly significant in both strains, P_tetlacZ^ΔM15 (P < 1 × 10⁻¹⁵) and P_tetlacZ^ΔM15ssrA (P < 2 × 10⁻³). While the observed growth deficiencies in the P_tetlacZ^ΔM15 strain may be caused by the presence of aggregates, our results do not exclude the possibility that expression of LacZ^ΔM15 may have deleterious consequences unrelated to protein aggregation. Further experiments will be needed to isolate the contribution of aggregate formation to cellular fitness (78).

Our LacZ-based model system to study proteotoxic stress phenotypes in E. coli may be used to explore and develop strategies for prevention or reversal of aggregate formation, which is important in a range of neurodegenerative diseases. A well-studied example of a tetrameric protein where assembly and aggregation are linked to disease states is the human protein transthyretin (TTR). TTR is implicated in familial amyloid diseases including neuropathy and cardiomyopathy (79). Known mutations that perturb TTR kinetic stability can lead to aggregation (80–82) and small molecules that stabilize the TTR tetramer are used therapeutically (79). Exploring the parallels and differences between LacZ and TTR assembly, aggregation, and degradation kinetics will deepen understanding of mechanisms underlying protein misfolding diseases (83).

In uncovering the mechanism of proteolytic stability of LacZ, we harnessed the power of ssrA tags, which are becoming a major tool in studying bacterial physiology and in synthetic biology applications (84, 85). Our in vivo quantification of multiple LacZssrA configurations along with our finding of transient degradation jointly expand the synthetic biology toolkit and point toward ways of modulating and controlling protein degradation. In general, proteins that tetramerize are not necessarily proteolytically stable, as demonstrated by several examples, including LacI (40, 86) and tryptophan synthase (87). Our findings point to the possibility of designing proteolytic stability via multimeric assembly, which may avoid having to redesign a protein fold. This indicates that the role of multimeric assemblies in proteolytic stability may be an important direction for future research.

Proteins in nature exhibit a wide range of proteolytic stabilities, from fully disordered proteins (88) to extremely stable ones such as LacZ. Proteolytically stable proteins can be inherited from mother to daughter cells over multiple generations providing long-term phenotypic adaptability. Transgenerational inheritance of LacZ is important in fluctuating environments (7), and understanding the role of proteolytic stability in shaping bacterial populations’ fitness may shed light on how gene expression levels are tuned for constant and fluctuating environments.

Materials and Methods

Details of Materials and Methods used in this paper, including E. coli strains and growth conditions; Plasmid and strain construction; LacZ activity measurements using the Miller assay; LacZ activity measurements during lac operon induction; LacZ degradation assay; Quantitative western blot; LacZ assay for M. florum—Lon degradation system; Growth dynamics and growth rate measurements; Microfluidics; Induction, repression, and α-complementation of LacZ^ΔM15; Microscopy; Quantification and statistical analysis are provided in SI Appendix, Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Data, Materials, and Software Availability

Data from Miller assays of expression kinetics can be found in Dataset S2. Microscopy images for quantification of aggregation and PI staining underlying Dataset S3 are available in the OSF repository (https://osf.io) with DOI: 10.17605/OSF.IO/CHSJM (89).